The Global Terrorism Database: how do researchers measure terrorism?

There are different approaches to measuring terrorism. Here is the approach of the Global Terrorism Database, a leading source of terrorism data.

Measuring terrorism worldwide helps us understand how people’s lives and livelihoods are affected by it.

But measuring terrorism attacks and deaths comes with many challenges. People do not always agree on what characteristics define an act of terrorism. Even once defined, these characteristics are difficult to assess, and challenging to distinguish from other forms of political violence and violent crime.

So how do researchers address these challenges and identify terrorist attacks and the deaths they cause?

What is the Global Terrorism Database?

In some of our work, we rely on data from the Global Terrorism Database (GTD).1

The database is maintained by the National Consortium for the Study of Terrorism and Responses to Terrorism (START) at the University of Maryland. Earlier phases of data collection were carried out by the private security agency Pinkerton (1970 to 1997), the Center for Terrorism and Intelligence Studies (1998 to 2008), and the Institute for the Study of Violent Groups (2008 to 2011).

START has received funding from various government agencies in the United States, United Kingdom, and Germany, but the data collection is conducted independently.

How does the Global Terrorism Database characterize terrorism?

GTD defines a terrorist attack as the threat or use of violence to achieve a political, economic, religious or social goal through intimidation or coercion by an actor that is not the state. More specifically, for an event to be considered terrorism:

Furthermore, at least two of the following three criteria must be met:

GTD does not require an attack to be successful to be included and does so as long as it is at least attempted.

Terrorism thereby differs from other types of violence. It is different from state repression because its perpetrators are not government actors. Its political, economic, religious, or social goals set it apart from violence for personal reasons, such as homicides or hate crimes. And its intent to intimidate and target civilians distinguishes it from other forms of armed conflict.

In practice, the lines between terrorism and other forms of violence can still be blurry. This is because the perpetrators of violence and their motivations are often difficult to identify.

Which deaths are included?

Deaths are only included if their number can at least broadly be identified. Both victims and perpetrators who died as a direct result of the attack are included.

What years and countries are covered?

GTD covers all countries worldwide, going back in time as far as 1970.

Because the data is based on news reports, attacks and deaths may be underreported in countries with sparser media, especially during early years.

How is terrorism measured?

How does GTD work to make its assessments valid?

GTD identifies terrorism events by having its researchers evaluate news reports, other contemporary sources, similar datasets, and academic research.4

Since 2012, the team has used keyword filters and natural language processing to identify which articles among the more than one million media articles published per day are relevant. They then review the 8,000 to 16,000 relevant articles per month to identify events and code their characteristics.

This review is done by six teams focusing on different characteristics (general information, location, perpetrators, targets, weapons and tactics, and casualties and consequences). The teams consist of several undergraduate and graduate students, supervised by a full-time research assistant.

GTD includes all events that meet its criteria for terrorism and are documented by at least one high-quality source.

For years since 1998, GTD allows filtering out any events that may not meet one’s own definition of terrorism by sharing disaggregated data on specific characteristics.

How does GTD work to make its assessments precise and reliable?

GTD uses many different types of sources: it primarily uses news reports, but also uses other contemporary sources, similar datasets, and academic research.

It prioritizes reporting from independent sources, those with a track record of providing reliable information, and prefers primary over secondary sources.

GTD also evaluates the news reports with a time lag and prioritizes sources published later to reduce the risk that incorrect information in the event’s immediate aftermath makes it into the database.

Fatalities and people injured are only included if their number can at least be broadly identified.

Their estimates are usually based on the most recent source if it is of high quality. If there are several most recent sources or the most recent source is not of high quality, then the estimate given in a majority of sources is used. If there is no majority estimate, the lowest estimate is used if it is of high quality.

And GTD works to identify events down to the city level and day to reduce the risk of counting an event several times.

How does GTD work to make its assessments comparable?

GTD also works to make its assessments comparable across countries and time.

It works to make its assessments comparable across countries by surveying English-language translations of media articles in more than 100 languages from around 150 countries.

Its current staff at START has also sought to make its definitions and methodologies consistent across earlier phases of data collection carried out by other institutions.

The availability of sources has still varied over time, and changes in 1998, 2008, and 2012 may in part be due to changes in the data collection. Data for 1993 is excluded entirely from the dataset because the original data was lost. START staff has been unable to comprehensively reconstruct the data.5

GTD acknowledges that high-quality data is harder to find in some countries than in others, which makes the estimates in such locations more conservative.

How are the remaining differences in the data dealt with?

GTD prioritizes reporting from independent sources, those with a track record of providing reliable information, primary over secondary sources, and sources published later.

If there are several most recent sources or the most recent source is not of high quality, then the deaths and injured estimates given in a majority of sources are used. If there is no majority estimate, the lowest estimate is used if it is of high quality.

Finally, GTD provides one indicator that describes its broad confidence in the data for each event.

How is the data made accessible and transparent?

GTD releases its data publicly and makes it straightforward to search on its website, download and use.

It publishes detailed descriptions of the questions and coding procedures. For events since 1998, it also includes its sources and a brief description of the incident.

How often and when is the data updated?

GTD typically releases a new version of the data each year.

We at Our World in Data aim to update our own data within a few weeks of the release.

What are the data’s shortcomings?

There are shortcomings in how GTD characterizes and measures terrorism.6

Because its coverage is primarily based on media reports, it is less likely to include events that received less media attention, such as events in remote locations, unsuccessful events, and those that were less deadly.

This means that trends especially in the early decades in countries beyond Western Europe and North America should be treated with caution. This is further compounded by GTD repeatedly improving its methodology, which increased the data’s quality, but reduced its comparability over time.

GTD has a broad understanding of terrorism and may include events that some people would not consider terrorist attacks. For example, its events do not have to have a political, economic, religious, or social goal, or have to target civilians, which some see as two core characteristics of terrorism.

What are the data’s strengths?

Despite these shortcomings, the data tells us a lot about how common terrorism is around the world, in the past and today.

GTD covers terrorism events worldwide, and comprehensively so in recent decades. Its data collection is ongoing and provides data on recent years.

Its inclusive definition of terrorism means that most events commonly considered terrorism are included, such as if they were domestic in nature7, only threatened violence (such as airplane hijackings) or merely used violence against property (such as bombings of infrastructure). And its fine-grained data allows its users to decide which events to include.

GTD also makes its data easy to access and search; and is free to use.

It is also not collected by any government, and therefore less susceptible to political pressure.

What is our summary assessment?

Whether GTD provides useful data on terrorism depends on the questions we want to answer.

The data may not give us satisfying answers if we are interested in more historical events in parts of the world less covered by the media, fine-grained changes over time, or specific types of terrorism.

In these cases, sometimes other data is available, such as the Database on Suicide Attacks8. But at other times, there simply is no better data.

But if we value high-quality data on broader global and historical patterns of terrorism, based on an inclusive and flexible understanding of what terrorism is, we can learn a lot from this data.

For these latter purposes, we use GTD in our reporting on terrorism.

Keep reading on Our World in Data


How common is terrorism? How does this differ across countries? And how is it changing over time? Explore research and data on terrorism.


I thank Edouard Mathieu for his very helpful comments and ideas about how to improve this article.


  1. START (National Consortium for the Study of Terrorism and Responses to Terrorism). 2021. Global Terrorism Database (GTD) [Data set]. University of Maryland.

  2. The economic goal must entail systemic economic change, not just profit.

  3. This is inferred from statements by the perpetrators before or after the attack, their past behavior, or the nature of the victims, weapons, or attack type.

  4. For more details, see: are: START (National Consortium for the Study of Terrorism and Responses to Terrorism). 2021. Global Terrorism Database codebook: Methodology, inclusion criteria, and variables. University of Maryland.

  5. You can find more information about why the data for 1993 is missing in GTD’s website FAQ

  6. This and the following section draw on three articles summarizing and reviewing some of the leading terrorism datasets:

    LaFree, Gary, and Laura Dugan. 2007. Introducing the Global Terrorism Database. Terrorism and Political Violence 19(2): 181-204.

    Sheehan, Ivan. 2012. Assessing and Comparing Data Sources for Terrorism Research. In: Lum, Cynthia, and Leslie Kennedy. Evidence-Based Counterterrorism Policy.

    Johnson, Stephen, and Gary Ackerman. 2023. Terrorism databases: Problems and Solutions. In: Frumkin, Lara, John Morrison, and Andrew Silke. A Research Agenda for Terrorism Studies.

  7. The alternative RAND database only covered transnational terrorism — where perpetrators and victims are from different countries — until the late 1990s.

  8. Pape, Robert, Alejandro Albanez Rivas, and Alexandra Chinchilla. 2021. Introducing the new CPOST dataset on suicide attacks 58(4): 826–838.

Cite this work

Our articles and data visualizations rely on work from many different people and organizations. When citing this article, please also cite the underlying data sources. This article can be cited as:

Bastian Herre (2023) - "The Global Terrorism Database: how do researchers measure terrorism?". Published online at Retrieved from: '' [Online Resource]

BibTeX citation

    author = {Bastian Herre},
    title = {The Global Terrorism Database: how do researchers measure terrorism?},
    journal = {Our World in Data},
    year = {2023},
    note = {}
Our World in Data logo

Reuse this work freely

All visualizations, data, and code produced by Our World in Data are completely open access under the Creative Commons BY license. You have the permission to use, distribute, and reproduce these in any medium, provided the source and authors are credited.

The data produced by third parties and made available by Our World in Data is subject to the license terms from the original third-party authors. We will always indicate the original source of the data in our documentation, so you should always check the license of any such third-party data before use and redistribution.

All of our charts can be embedded in any site.